Pedagogical AI-based architecture for encouraging self-regulated learning behavior in students

Abstract

With the advancement of educational technologies, Virtual Learning Environments have become essential for promoting new teaching and learning methods, especially in distance education and hybrid contexts. These environments allow students to access content, complete activities, and interact with peers and instructors in a flexible and personalized manner. In this scenario, Self-Regulated Learning stands out as a key com- petency, as it enables learners to autonomously manage, monitor, and direct their own learning process. This study proposes and validates an Artificial Intelligence - supported Pedagogical Architecture to foster SRL in VLEs, aiming to enhance students’ autonomy and engagement. Initially, a systematic literature review was conducted, which identified research gaps and guided the PA design. Subsequently, Proofs of Concept were carried out using data from the Open University Learning Analytics Dataset and from Moodle at IFSULDEMINAS – Campus Carmo de Minas, applying Educational Data Mining tech- niques and clustering algorithms. These analyzes allowed the identification of behavioral patterns, SRL profiles, and significant correlations between engagement and academic performance. In the final stage, the PA was implemented and evaluated in the context of an online Introduction to Python Programming course. Among the resources integrated into the VLE, the Time Tracker SRL plugin stands out, developed to monitor the time dedicated to learning activities and provide automated feedback. Other plugins, such as Configure Reports, Completion Progress, Analytics Graphs, and OpenAI Chat, were also employed to support the self-regulation process. The results showed that the PA had a significant impact in promoting SRL, with a positive correlation between engagement and academic performance. The triangulation of evidence—based on VLE log analysis, self-regulation questionnaires, and focus group interviews—confirmed the effectiveness of the PA, validating its potential to develop SRL skills, foster autonomy, and improve stu- dent performance. Thus, the proposed approach constitutes an innovative, scalable, and adaptable solution to support and personalize learning in VLEs.